Monarcha Turns Old Maps Into Data Because Nobody Wants to Trace Property Lines by Hand

The Macro: The GIS Industry Runs on Manual Labor and Nobody Talks About It

There is a strange contradiction in the geospatial industry. The tools have gotten incredibly sophisticated. ESRI’s ArcGIS can do things that would have been science fiction twenty years ago. QGIS gives you open-source power that rivals the expensive stuff. Planet Labs is shooting satellite imagery of the entire Earth every single day. The visualization layer is futuristic.

But the data entry layer is stuck in 1997.

If you work at a mining company, a civil engineering firm, or a land management agency, a huge portion of your spatial data exists as scanned documents. Paper maps that got photographed. Plats from county recorders. Survey documents from the 1960s. Geological reports with hand-drawn cross sections. These documents contain valuable spatial information, but it is trapped in pixels. You cannot query a JPEG. You cannot overlay a PDF onto a coordinate system. To turn these documents into usable GIS data, someone has to sit down and manually trace every boundary, plot every point, and type in every label.

This is called georeferencing and digitization, and it is one of the most tedious jobs in the geospatial world. Companies pay GIS technicians to do this work for hours and hours. Large organizations have entire teams dedicated to it. The backlog is always growing because new documents come in faster than humans can process them.

The market for this is not small. Mining exploration alone generates mountains of historical map data that needs to be digitized for modern analysis. Infrastructure projects require converting decades of as-built drawings into queryable spatial databases. Land title companies need to verify boundaries from old survey documents. Every one of these workflows involves someone staring at a scanned image and manually recreating it in GIS software.

Esri has some automated georeferencing tools but they work best on clean, standardized documents. The moment you throw a wrinkled survey from 1978 at them, accuracy falls apart. Trimble and Hexagon have digitization workflows but they still require significant manual intervention. Nobody has cracked the “throw any document at it and get good data back” problem.

The Micro: Ninety Seconds From Scanned Map to Spatial Dataset

Monarcha is an AI-powered georeferencing platform that converts scanned maps, plats, surveys, and land documents into queryable spatial data. You upload a document. The vision AI processes it. Ninety seconds later, you have a georeferenced GIS layer with coordinates, boundaries, and metadata that you can drop into any spatial analysis workflow.

James Spokes and Everett Lee are the cofounders, running the company out of San Francisco as part of Y Combinator’s Summer 2025 batch. The founding story maps directly to the problem. Anyone who has worked in mining exploration or land management has experienced the pain of sitting on a pile of historical maps that contain exactly the data you need but cannot use because it is not in a queryable format.

The technical challenge here is genuine. Scanned maps are messy. They have different projections, scales, orientations, and levels of degradation. A survey from 1955 looks nothing like a satellite image from 2024. The text is sometimes handwritten. The boundaries are sometimes drawn with a marker that bled through the paper. Getting a vision model to reliably extract spatial information from this kind of input and then accurately place it on a coordinate grid is not a trivial computer vision problem.

What I find compelling about Monarcha’s approach is the conversational interface layer. Once your documents are georeferenced, you can ask questions in natural language. “Show me all parcels within two miles of this river.” “Overlay wildfire risk data on these mining claims.” “Which of these survey boundaries overlap with the proposed pipeline route.” That turns a static data conversion tool into an analysis platform, which is where the real stickiness lives. Converting a map is a one-time task. Querying your spatial data is an ongoing workflow.

The go-to-market is clearly enterprise. Mining companies, civil engineering firms, and land intelligence organizations are the target. No public pricing, enterprise sales calls required. A demo environment exists at demo.monarcha.ai, which suggests the product is functional enough to show rather than just describe.

The competitive moat here is accuracy. If Monarcha can reliably georeference documents that ArcGIS cannot handle automatically, that is a defensible position. The document types are varied enough that training data is a real barrier. Every mining company’s historical maps look different, and building a model that generalizes across that variety takes time and data that competitors do not have yet.

The Verdict

I think Monarcha is solving a real problem that the established GIS vendors have mostly ignored because it is not glamorous. Esri wants to sell you a platform. Monarcha wants to solve the boring bottleneck that prevents you from using the platform effectively. That is often where the most durable businesses get built.

The question is whether the accuracy holds up across the wild variety of document types that exist in the real world. A demo is one thing. Processing ten thousand documents from a mining company’s filing cabinet, with varying quality, formats, and ages, is another thing entirely. The ninety-second processing time claim is impressive if it is real and consistent.

In thirty days, I want to see case studies from actual mining or engineering clients. Not hypothetical use cases but real documents processed with real accuracy metrics. In sixty days, the question is whether the natural language query interface is actually useful or whether GIS professionals default back to their existing tools after the initial conversion. In ninety days, I want to know the repeat usage pattern. If clients upload a batch of documents once and never come back, this is a services business disguised as a platform. If they are querying their spatial data weekly, Monarcha has something sticky.